How m-awais-11/datapilot-ai works
This is an AI data-science application: users upload datasets, launch experiments, follow an autonomous analysis and modeling workflow, then review results, explanations, and reports or chat about an experiment. The README documents the ordered agent stages; sampled task code confirms that experiment execution invokes the LangGraph workflow, and sampled frontend code confirms API access and pipeline-result polling. Several API handlers and agent implementations were not sampled, so uncertain internal service wiring is omitted.
Main entry points: api.service.ts API client · supervisor.py Workflow supervisor · app.component.ts Angular app · connection.py Application database
Components 31
Web application
- Angular app
frontend/src/app/app.component.ts - Sign-in
frontend/src/app/auth/login.component.ts - Dataset upload
frontend/src/app/dashboard/dashboard.component.ts - Experiment setup
frontend/src/app/experiments/new-experiment.component.ts - Pipeline results
frontend/src/app/experiments/pipeline.component.ts - Prediction form
frontend/src/app/predictions/prediction-form.component.ts - Chat interface
frontend/src/app/chat/chat.component.ts - API client
frontend/src/app/core/services/api.service.ts
Backend API
- Authentication API
backend/app/api/auth.py - Dataset API
backend/app/api/datasets.py - Experiment API
backend/app/api/experiments.py - Prediction API
backend/app/api/predictions.py - Reports API
backend/app/api/reports.py - Chat API
backend/app/api/chat.py
Experiment workflow
- Experiment task
backend/app/tasks/experiment_tasks.py - Workflow supervisor
backend/app/agents/supervisor.py - Data understanding
backend/app/agents/data_understanding.py - Data cleaning
backend/app/agents/cleaning.py - Exploratory analysis
backend/app/agents/eda.py - Feature engineering
backend/app/agents/feature_engineering.py - Model selection
backend/app/agents/model_selection.py - Model training
backend/app/agents/training.py - Model evaluation
backend/app/agents/evaluation.py - SHAP explainability
backend/app/agents/explainability.py - Report generation
backend/app/agents/reporting.py
Persistence and artifacts
- Application database
backend/app/database/connection.py - Dataset service
backend/app/services/dataset_service.py - MLflow tracking
backend/app/services/mlflow_service.py
Serving and assistance
- Grounded chat service
backend/app/services/chat_service.py - LLM provider
Other components
- User
Connections 39
- User → Angular app: uses
- Angular app → Sign-in: routes to
- Angular app → Dataset upload: routes to
- Angular app → Experiment setup: routes to
- Angular app → Pipeline results: routes to
- Angular app → Prediction form: routes to
- Angular app → Chat interface: routes to
- Sign-in → API client: requests auth
- Dataset upload → API client: uploads data
- Experiment setup → API client: creates run
- Pipeline results → API client: polls results
- Prediction form → API client: requests prediction
- Chat interface → API client: sends messages
- API client → Authentication API: HTTP requests
- API client → Dataset API: HTTP requests
- API client → Experiment API: HTTP requests
- API client → Prediction API: HTTP requests
- API client → Reports API: HTTP requests
- API client → Chat API: HTTP requests
- Dataset API → Dataset service: dataset operations
- Dataset service → Application database: persists metadata
- Experiment API → Experiment task: starts experiment
- Experiment task → Application database: loads and updates
- Experiment task → Workflow supervisor: invokes workflow
- Workflow supervisor → Data understanding: dispatches stage
- Workflow supervisor → Data cleaning: dispatches stage
- Workflow supervisor → Exploratory analysis: dispatches stage
- Workflow supervisor → Feature engineering: dispatches stage
- Workflow supervisor → Model selection: dispatches stage
- Workflow supervisor → Model training: dispatches stage
- Workflow supervisor → Model evaluation: dispatches stage
- Workflow supervisor → SHAP explainability: dispatches stage
- Workflow supervisor → Report generation: dispatches stage
- Model training → MLflow tracking: logs metrics
- Chat API → Grounded chat service: handles conversation
- Grounded chat service → Application database: reads context and saves
- Grounded chat service → LLM provider: requests answers
- Reports API → Application database: retrieves report records
- Prediction API → Application database: retrieves model records